Great work, and thanks for sharing the code with us.
I noticed that you had added attacking images for both LLaVA and InstructBLIP in the updated paper, however, it seems that you didn't share those images in this repo. I'm wondering if it's possible to share those attacking images, so we could investigate more without launching the training process. Thanks in advance!
Great work, and thanks for sharing the code with us.
I noticed that you had added attacking images for both LLaVA and InstructBLIP in the updated paper, however, it seems that you didn't share those images in this repo. I'm wondering if it's possible to share those attacking images, so we could investigate more without launching the training process. Thanks in advance!